Yemen airstrike database project
Partnering with GLAN, Bellingcat, Mnemonic/Yemeni Archive, VFRAME, Huridocs, and Digital Evidence Vault, with funding from Nesta’s Collective Intelligence Grants scheme, we developed a database of airstrikes & cluster bomb attacks in Yemen. This user-friendly system enables the cross-referencing of large quantities of evidence and will be available to partner organisations seeking to promote accountability in the future. It brings together key pre-existing technologies to allow online user generated content and private evidence to be stored and viewed together. VFRAME developed computer vision algorithms for the detection and categorisation of videos and pictures that contain indicators of cluster munitions. This research was featured in the MIT Tech Review. You can read more about the project here.
Building a Secure System for Reporting Human Rights Violations and Receiving Psychosocial Support
In partnership with Ateneo de Manila University, The Philippines and WAPR, we created a secured incident monitoring platform that allows for submission or recording of victim/witness reports via SMS or Web. The system is designed for localized and customized use in communities who need an ICT-based Human Rights Violation reporting tool which allows for verification and validation. Features of the system including its ability to detect emotions and behaviours from narratives through natural language processing, which in turn can signpost users to psychosocial support available, and the incorporation of geospatial elements so that incidents can be viewed on a map. This project received financial support from Swansea University’s GCRF and Cherish-DE research funds.
Developing the Knowledge Hub Framework – a tool for open source human rights investigations
With funding from HEFCW’s Research Wales Innovation Fund, we are currently developing the existing KHF into a web-based set of tools. The existing Virtual Machine format, which runs locally on users’ machines, works very well for protecting information, but because of its size, it requires some technical knowhow and a fairly large operating system to run, so it is not ideal for many human rights investigators. The web-based platform would allow trusted users to carry out verification tasks without first having to download the VM.
Supporting the Identification of Social Media Hate Crimes using Natural Language Processing
With support from the University of Manchester’s ESRC Impact Acceleration Account, we will package the NLP methods for hate speech detection developed as part of OSR4Rights into a graphical web-based tool that can be used by human rights investigators to identify hate speech.